Decision making and principled exploitation
13.1 Optimize the distribution, not the anecdote
A natural beginner question is "What should I do with this hand?" A more advanced question is "What does my strategy do with the set of hands that reach this node?" If every missed draw bluffs, the bluff frequency may become excessive. If only the absolute nuts bet large, observant opponents can fold too much. Your individual action inherits constraints from the rest of your range.
13.2 Marginal exploitation
Suppose an opponent calls river bets too often. The direct exploit is not "bluff never" in every situation. It is to expand thin value betting and contract marginal bluffs where the evidence applies. If an opponent folds too much, expand profitable bluffs and sometimes reduce the weakest value bets that no longer receive enough calls. If an opponent underbluffs, tighten marginal bluff catches.
| Observed deviation | Directional response in matching spots |
|---|---|
| Calls too often | Value bet thinner; bluff less at the margin. |
| Folds too often | Bluff more; avoid overvaluing thin calls from them. |
| Bluffs too often | Call wider with bluff catchers. |
| Bluffs too little | Fold more marginal bluff catchers. |
| Raises only very strong hands | Continue tighter versus raises until evidence changes. |
| Uses size as a strong tell | Condition ranges on size, but demand a real sample. |
13.3 Bayes, but with humility
A live read, timing tell, or population tendency is evidence about hidden state. Formally it modifies likelihood ratios. Practically, these signals are noisy, selected, and vulnerable to confirmation bias. The correct mental move is a probability update, not a declaration: "This size increases the weight on strong hands" is better reasoning than "He always has it."
13.4 Exploit at the margins
A useful principle from the decision-making lecture is to exploit at the margins. Keep the central structure of a robust strategy and shift the borderline hands first. This reduces the cost of being wrong. If you think someone over-folds by a modest amount, add the best candidate bluffs before turning every hand into a bluff. If you think they under-bluff, fold the weakest bluff catchers before folding hands that dominate their value mistakes.
Model-risk rule
The more uncertain your opponent model, the smaller the exploitative deviation should be. Confidence should control step size.
13.5 Know thyself
Your own tendencies are part of the game. Fatigue can narrow attention; recent losses can increase risk-seeking; recent wins can create overconfidence; time pressure can force defaults. A review database is valuable partly because it reveals systematic personal deviations that subjective memory edits away.